Interesting changes in HDD tech

[x88x]

A very well written article on the limitations of current HDD tech (particularly HDD sectors), and the changes that have been happening over the last few years to try and combat them.

http://arstechnica.com/microsoft/news/2010/03/why-new-hard-disks-might-not-be-much-fun-for-xp-users.ars/

[SXRguyinMA]

interesting

[Xpirate]

I read about that earlier this week. XP will become as obsolete as Windows 98. It's only a matter of time.

[artoodeeto]

fascinating...wonder if it applies equally to SSDs vs traditional platters, since SSDs (I think anyway, I could be wrong) really do just store 1's and 0's...

[x88x]

I think most SSDs are already built with 4096 bit sectors. I could be wrong about that though.

[artoodeeto]

actually my question is more, are SSDs prone to similar instances/numbers of errors the way the article describes the traditional platter drives as being? ie, are SSDs as inaccurate at storing information as a platter hard drive, to the point that similar amounts of data reconstruction must be performed each time the computer accesses the disk? Not that it really matters in terms of practicality, I love my SSD and it's working great, I'm just curious.

[crenn]

SSDs store data different to platter based hard drives. I don't think SSDs would have the same problems with tiny magnetic fluxes.

[x88x]

SSDs use NAND flash (http://en.wikipedia.org/wiki/NAND_flash) memory chips for storage, which operate radically differently than HDD platters do. Basically, magnetic HDD platters use the polarity of the magnetic charge over a portion of the disk to store a bit, whereas NAND flash uses the position of a transistor. So unless something else affects them that I'm not aware of, SSDs should not have to do all the error correction that HDDs do. However, MLC SSDs do have to do load-balancing to extend the life of the drive to a reasonable level, so you do still have overhead..just a different kind of overhead.

[chaksq]

I read through the article, very interesting. I get that the new standards will increase maximum storage capacities of drives. I'm a little confused though, will the larger sector sizes speed up drive seek times or is this change solely for increasing drive overall sizes?

[x88x]

I'm a little confused though, will the larger sector sizes speed up drive seek times or is this change solely for increasing drive overall sizes?

Technically, neither, but the end result of the change is that it will enable larger capacities. The source of the problem is that data density is increasing faster than our abilities to accurately measure an appropriately small arc of the platter. Because of this, more error-correction area is needed per sector, but with 512b sectors, increasing the error-correction space to what would be needed to ensure the same level of data loss (I think it was something around 1 bit per 10^14) ended up raising the required overhead capacity to such a point that the change in drive capacity was less than the change in overhead needed (resulting in a negative change in usable storage space). I believe the number given for the new level of error correction was 63% overhead. By switching to 4096b sectors, they were able to drastically reduce the amount of overhead required, enabling a positive change in usable storage space for higher capacity drives while still retaining the same level of data loss.

I don't think that seek times were mentioned, though I could be misremembering that. I don't know that seek times would actually be affected much at all.